[JCE]Fast & stable CN/v8/Heavy/Tube/XHV miner, CPU+GPU, Vega56 1800+ RX580 1200+

[JCE-Miner]

Turtle:

CN_Turtle Testnet – Now that PoW algo [1] has been announced for block 1,200,000 lets start testing it all out. The project needs some people to run some infrastructure (seed nodes, pool, explorer). We’ll also need some testers to send a few hundred hashes at it and check the miners are all OK as they get updated. In the next day or two we’ll be able to compile the testnet bits [2] and even just run straight from containers [3] then we’ll aim for an upgrade height for around midnight Saturday UTC specifically block 1,060,600. So anyone that wants to help out with infrastructure bits please ping me SoreGums#8071 in discord [4] to get added to #dev_testnet, everyone has access to view this channel if you’re curious as well. – soregums

Sure, i'm waiting for their test pool.
Source: https://blog.turtlecoin.lol/archives/this-week-last-week-in-turtlecoin-dec-10th-17th-2018/

@migo:
Looks like my miner uses 8 threads and stak uses 16, and stak is probably right to do such.
Can you copy-paste the beginning of my log, where it detects the CPUs ?

Otherwise, try this manual config, which is very like the one of stak:

Code:

"cpu_threads_conf" : 
[ 
     { "cpu_architecture" : "auto", "affine_to_cpu" : 0, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 1, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 2, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 3, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 4, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 5, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 6, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 7, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 8, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 9, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 10, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 11, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 12, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 13, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 14, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 15, "use_cache" : true },
]

@Unclwish:
try those:

Code:

"cpu_threads_conf" : 
[ 
     { "cpu_architecture" : "auto", "affine_to_cpu" : 0, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 1, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 2, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 3, "use_cache" : true },
]

Code:

"cpu_threads_conf" : 
[ 
     { "cpu_architecture" : "auto", "affine_to_cpu" : 0, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 1, "use_cache" : false },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 2, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 3, "use_cache" : false },
]

But on a small AES intel CPU, i'm not that faster, when i am. Except on Bittube. It's possible i don't even compensate for the 1.5% fees, assuming you compiled xmrig with no fees.

[1715319304]

1715319304

[1715319304]

1715319304

[UnclWish]

@Unclwish:
try those:

Code:

"cpu_threads_conf" : 
[ 
     { "cpu_architecture" : "auto", "affine_to_cpu" : 0, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 1, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 2, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 3, "use_cache" : true },
]

Code:

"cpu_threads_conf" : 
[ 
     { "cpu_architecture" : "auto", "affine_to_cpu" : 0, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 1, "use_cache" : false },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 2, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 3, "use_cache" : false },
]

But on a small AES intel CPU, i'm not that faster, when i am. Except on Bittube. It's possible i don't even compensate for the 1.5% fees, assuming you compiled xmrig with no fees.

1st variant - 57 h/s
2nd - 54 h/s

My variant

Code:

"cpu_threads_conf" : 
[ 
     { "cpu_architecture" : "auto", "affine_to_cpu" : false, "use_cache" : true, "multi_hash" : 5 },
     { "cpu_architecture" : "auto", "affine_to_cpu" : false, "use_cache" : true, "multi_hash" : 5 },
     { "cpu_architecture" : "auto", "affine_to_cpu" : false, "use_cache" : true, "multi_hash" : 5 },
]

This variant gives 64 h/s and CPU load only 75% for comfort doing my work)))) On XMRig tha same config gives 72-75 h/s (and you're right about fee)))
Overloading L3 cache by multi_hash gives additional speed...

[migo77]

@migo:
Looks like my miner uses 8 threads and stak uses 16, and stak is probably right to do such.
Can you copy-paste the beginning of my log, where it detects the CPUs ?

Otherwise, try this manual config, which is very like the one of stak:

Code:

"cpu_threads_conf" : 
[ 
     { "cpu_architecture" : "auto", "affine_to_cpu" : 0, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 1, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 2, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 3, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 4, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 5, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 6, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 7, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 8, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 9, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 10, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 11, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 12, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 13, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 14, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 15, "use_cache" : true },
]

Hi, It is my manual config for xmr-stak which gives me max hr. For your miner it is  not best one, I'll post results later. I've done hours of config fine tuning to get max hr already. I must take care of my son now, sorry.

[JCE-Miner]

Core i3: i admit defeat versus xmrig in that case, those low-cache CPUs love to have their registers and cache flooded and i cannot do any theorical optim without having one, and i haven't. Even on my optimized asm for big Intel i barely give better perf, and it's still harder against a no-fee version of xmrig.

The dual Xeon Nehalem is more interresting, while i give slightly better perf on 8 threads that stak on 16, i still expected a far bigger difference. I've like +80% on core2, that's a surprise to get +1% on nehalem. But again, i've none of those CPU so hard to tell why
I got reports that on Opteron (old non-aes AMD) i also give +30% perf, so maybe the Nehalem is a special case of non-AES cpu which runs bad on my code

I'm polishing the remote-manageable version of JCE, with an ultra-ultra-simplified set of commands that can be triggered from a browser.
I'll write a dedicated doc, but the principle will be to navigate to some URLs to send commands.
Assuming the HTTP server is enabled on address righost:1234, navigate to:

http://righost:1234 to get JSON status, as before
http://righost:1234/pause to pause all
http://righost:1234/resume to resume all
http://righost:1234/pause-cpu to pause all CPUs
http://righost:1234/pause-gpu to pause all GPUs
http://righost:1234/pause-gpu-N to pause GPU N (N decimal or hexa)
http://righost:1234/stop to kill the miner
http://righost:1234/restart to restart the miner

The last command allows you to change the config files in-between, so after the restart, the miner can mine a different coin, on a different pool, with more or less CPU/GPU etc.
All commands are logged, even if it's non-sense, for security. Like pausing a GPU you don't have, or resuming a miner that's already running. In such case, it's a no-op, except for the log itself.

[pp55]

I confirm the same hashes with 0.33g,h,k CPU versions on AMD FX-8320E, CNv8.

310-315 h/s with config:
--auto -t 8
or the same

Code:

"cpu_threads_conf" : 
[ 
     { "cpu_architecture" : "auto", "affine_to_cpu" : 0, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 1, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 2, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 3, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 4, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 5, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 6, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 7, "use_cache" : true },
]

Any better configs for AMD FX-8320E, CNv8?

[JCE-Miner]

8 cores, no Hyperthread-like, 8M L2 + 8M L3 exclusive, so i think that's the best config possible.

edit: the remote managment doc is ready
https://github.com/jceminer/cn_cpu_miner/blob/master/README.md#remote-managment

The release itself comes soon, i need to test it more.

[migo77]

@migo:
Looks like my miner uses 8 threads and stak uses 16, and stak is probably right to do such.
Can you copy-paste the beginning of my log, where it detects the CPUs ?

Otherwise, try this manual config, which is very like the one of stak:

Code:

"cpu_threads_conf" : 
[ 
     { "cpu_architecture" : "auto", "affine_to_cpu" : 0, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 1, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 2, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 3, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 4, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 5, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 6, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 7, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 8, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 9, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 10, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 11, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 12, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 13, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 14, "use_cache" : true },
     { "cpu_architecture" : "auto", "affine_to_cpu" : 15, "use_cache" : true },
]

Hi JCE, as I said, this config is good for xmr-stak not for jce:
{
  "hashrate":
  {
    "thread_0": 9.05,
    "thread_1": 8.68,
    "thread_2": 8.66,
    "thread_3": 8.68,
    "thread_4": 8.80,
    "thread_5": 8.81,
    "thread_6": 8.80,
    "thread_7": 8.80,
    "thread_8": 9.08,
    "thread_9": 8.69,
    "thread_10": 8.66,
    "thread_11": 8.67,
    "thread_12": 8.80,
    "thread_13": 8.80,
    "thread_14": 8.81,
    "thread_15": 8.80,
    "thread_all": [9.05, 8.68, 8.66, 8.68, 8.80, 8.81, 8.80, 8.80, 9.08, 8.69, 8.66, 8.67, 8.80, 8.80, 8.81, 8.80],
    "total": 140.51,
    "max": 140.58
  },
  "result":
  {
     "wallet": "xxx",
     "pool": "xxx",
     "ssl": false,
     "reconnections": 0,
     "currency": "Monero (XMR/XMV)",
     "difficulty": 6240,
     "shares": 4,
     "hashes": 36240,
     "uptime": "0:06:16",
     "effective": 96.38
  },
  "miner":
  {
     "version": "jce/0.33k/cpu",
     "platform": "Dual Intel(R) Xeon(R) CPU E5520 @ 2.27GHz",
     "system": "Linux 64-bits",
     "algorithm": "15"
  }
}

This is with Nehalem architecture optimizations, but when I manually set generic_sse4 architecture I get +1 H/s what is interesting:

{
  "hashrate":
  {
    "thread_0": 19.16,
    "thread_1": 19.16,
    "thread_2": 19.16,
    "thread_3": 19.16,
    "thread_4": 19.21,
    "thread_5": 19.21,
    "thread_6": 19.21,
    "thread_7": 19.21,
    "thread_all": [19.16, 19.16, 19.16, 19.16, 19.21, 19.21, 19.21, 19.21],
    "total": 153.46,
    "max": 153.46
  },
  "result":
  {
     "wallet": "xxx",
     "pool": "xxx",
     "ssl": false,
     "reconnections": 0,
     "currency": "Monero (XMR/XMV)",
     "difficulty": 6660,
     "shares": 2,
     "hashes": 19990,
     "uptime": "0:01:50",
     "effective": 181.73
  },
  "miner":
  {
     "version": "jce/0.33k/cpu",
     "platform": "Dual Intel(R) Xeon(R) CPU E5520 @ 2.27GHz",
     "system": "Linux 64-bits",
     "algorithm": "15"
  }
}

I think there is something wrong in your optimizations for this architecture... I can sell you one Sun 1U server for 3XMR + shipping  for development if interested.

[migo77]

I got reports that on Opteron (old non-aes AMD) i also give +30% perf, so maybe the Nehalem is a special case of non-AES cpu which runs bad on my code

Yes I can confirm this too, with xmr-stak I get 160H/s and with JCE:

{
  "hashrate":
  {
    "thread_0": 14.55,
    "thread_1": 14.55,
    "thread_2": 14.56,
    "thread_3": 14.55,
    "thread_4": 14.55,
    "thread_5": 14.56,
    "thread_6": 14.64,
    "thread_7": 14.64,
    "thread_8": 14.64,
    "thread_9": 14.64,
    "thread_10": 14.64,
    "thread_11": 14.64,
    "thread_all": [14.55, 14.55, 14.56, 14.55, 14.55, 14.56, 14.64, 14.64, 14.64, 14.64, 14.64, 14.64],
    "total": 175.11,
    "max": 176.21
  },
  "result":
  {
     "wallet": "xxx",
     "pool": "xxx",
     "ssl": false,
     "reconnections": 0,
     "currency": "Monero (XMR/XMV)",
     "difficulty": 5160,
     "shares": 2673,
     "hashes": 13941780,
     "uptime": "22:27:24",
     "effective": 172.45
  },
  "miner":
  {
     "version": "jce/0.33k/cpu",
     "platform": "Dual Six-Core AMD Opteron(tm) Processor 2435",
     "system": "Linux 64-bits",
     "algorithm": "15"
  }
}

[riskyfire]

I've uploaded a video to help beginners set custom cpu cores & threads.

JCEMiner Cryptonote CPU Miner - How to Set Custom No. CPU Core & Thread Count

[Iamtutut]

JCE, about MASARI Fork:

"Barring any setbacks, on January 11, 2019 Masari will fork to include SECOR (Uncle Mining) + Bulletproofs, & move to the next generation of CN-Fast for its proof of work algorithm.
CN-Fast v2 moves CN-Fast from CNv7 to CNv8, and continues with halving iterations. Open source mining binaries will be provided on the Masari GitHub sometime before the fork date."

[UnclWish]

JCE, what plans you have now? On GPU part... Any enchancements for speed, or energy saving?

[ganzocrypt]

any chance to have a version for macOS ? got a dual Xeon 20 cores Hackintosh.

[JCE-Miner]

@migo77 : so you also have a Dual Opteron?
Are you an old hardware collector? Those servers are getting rarer.
I don't plan to buy one now, the old cpu mining is evaporating because of the low coin prices, i entered quite late into the miner dev game
Even your Opteron perf are quite disapointing, that's like +8% here, but the +30% was for v7, i didn't rebench meanwhile.

@ILYA_Zzz:
Here's your remote managment. But it's not the same as Claymore, i made an ultra-simplified one.

Mac: nope sorry, i'm very very not a mac person. But aren't Macs just rebranded x64 PCs now?
While my linux users are like 2%, i cannot imagine how few Mac users I could have

Planned: forks. I've tons of forks to add, and don't forget i've 4 cpu versions to update, not counting the multi-things (aes/not, 32/64). It requires a lot of time.

At least here's a new one:

JCE CPU Windows 0.33m

Turtle v2 fork
Dego coin
HTTP Remote managment

Still have to add the new Masari and Stellite, while i'm a lot less motivated to add Hycon and Webchain.

[Iamtutut]

Forks for known crypto should be the priority, both CPU and GPU version. I'd like to be able to mine TRTL with GPU again, right now it's a waste because of FPGAs and / or nicehash.

[migo77]

@migo77 : so you also have a Dual Opteron?
Are you an old hardware collector? Those servers are getting rarer.
I don't plan to buy one now, the old cpu mining is evaporating because of the low coin prices, i entered quite late into the miner dev game
Even your Opteron perf are quite disapointing, that's like +8% here, but the +30% was for v7, i didn't rebench meanwhile.

Hi JCE, yes I've only one dual opteron, all other CPU miners are old xeons.  I've some 270x,280x,290x and vega64 too, but all on Linux so no chance for JCE.

Yes, V8 fork degraded performance for all my old servers by 30% and more sometimes using XMR-stak. ( Som of looses can be recovered with JCEminer but not all. Most problematic is Xeon E5520 and I've several of them

Yes, collector of old good HW mostly SUN servers

[JCE-Miner]

Fine, I had been such a collector, and still use old rigs like Core2 + HD7850, but they're offline now, the coin prices don't cover the electricity cost any longer.
My preferred retro PC i built was a K6-3+ on a P5A from 1998 with some crazy hardware attached to it, like a GeForce4, 1G ram and RAID disks. I wanted to make the most powerful Socket7 PC ever, and i was probably not far
Sad the 32-bits only no longer let me use it for real. It's not a matter of power, but of software compatibility.

I've a working Turtle-v2 capable GPU version but i'm not satisfied with the performance yet. And i want to make the remote restart working fine, which is harder than for just CPU mining.

[ganzocrypt]

any configuration for AMD RX 580 8Gb gpus?
want to try a config.txt instead of auto

thx

[JCE-Miner]

config:
https://github.com/jceminer/cn_gpu_miner#configuration

here you have some good config, but fine tuning may still depend on your exact card.

I'm done with Turtle v2 GPU, and satisfied with the perfs. The only thing i still need to test more is the GPU restart (an unrelated feature).

One of my RX560 gave the exact same speed as my Ryzen 1600 stock for Monero v8: 480 h/s.
On Turtle v2, the RX jumped to 4150, and the Ryzen to 9300
really a good algo for CPU

I've uploaded a video to help beginners set custom cpu cores & threads.

I noticed it a few days ago, thanks again!

[ganzocrypt]

I saw those configuration and tried some.

I am mining uPlexa and did not see difference between --auto and using the following for RX580 8Gb and RX580 4G

{ "mode" : "GPU", "worksize" : 8, "alpha" : 64, "beta" : 8, "gamma" : 4, "delta" : 4, "epsilon" : 4, "zeta":4, "index" : 1, "multi_hash":1696 },
{ "mode" : "GPU", "worksize" : 8, "alpha" : 64, "beta" : 8, "gamma" : 4, "delta" : 4, "epsilon" : 4, "zeta":4, "index" : 1, "multi_hash":1696 },
{ "mode" : "GPU", "worksize" : 8, "alpha" : 64, "beta" : 8, "gamma" : 4, "delta" : 4, "epsilon" : 4, "zeta":4, "index" : 2, "multi_hash":992 },
{ "mode" : "GPU", "worksize" : 8, "alpha" : 64, "beta" : 8, "gamma" : 4, "delta" : 4, "epsilon" : 4, "zeta":4, "index" : 2, "multi_hash":992 },

\With auto on 6 RX580 8Gb + 1 RX 580 4Gb I am doing almost 22kh/s

I asked to see if there was some more optimization I could try.
Thx

[JCE-Miner]

try beta=16 (sometimes it's better) otherwise increase/decrease by step of 16 the multi_hash
you can make the two threads of the same GPU different in case it's better.

uPlexa is very like the old Turtle v1, you should start with the reference v7 config and then tweak, i didn't have time to search for the best configs for this algo, the auto config applies the same values as for the Turtle v1.
A 4G card should be able to go much higher than 992 on uPlexa, since it uses 1M per multi_hash.

edit:
the 0.33b15 GPU is online:

Same changes as for the 0.33m CPU:
* Turtle v2
* Remote Managment
* Coin: DeroGold
+ IGP extra fix for autoconfig

the autoconfig for Turtle v2 will apply the same params as for Monero, since i've really no time to test all cases.